WebLogic Server Work Managers and Overload Protection
Download as PPTX, PDF18 likes15,095 views
The document provides an overview of WebLogic Server Work Managers, which are designed to optimize resource allocation and manage workload effectively without complex manual configurations. Key features include dynamic thread pool management, prioritization of application requests, and overload protection mechanisms that adapt based on resources and application-defined goals. Examples illustrate how applications can specify resource requirements, ensuring efficient handling of requests while protecting against resource hogging and cascading failures.
WebLogic Server Work Managers and Overload Protection
- 2. WebLogic Server Work ManagersJames BayerWebLogic Server Product Management
- 3. AgendaWebLogic Server Work Manager OverviewFeature Demonstration
- 4. <Insert Picture Here>WLS Work ManagerOverview
- 5. Real World Overload Protection
- 6. WebLogic Server Overload ProtectionExamples where this is applicableProtect against cascading failures
- 7. If maximum number of Database connections are in use, do not allocate new threads to service web requests that need a database connection
- 8. Protect against hogging applications
- 9. If multiple applications are deployed to the same server, ensure badly behaving applications do not consume all resourcesWork Management ConcernsOptimally tuning servers is hard!Thread pool size for optimum performance?Workload can vary (time of day, event driven, etc.)Handle overloadconditions gracefullyPrioritization of workAcross modules within single applicationAcross multiple applicationsAcross multiple classes of usersOrdered processing of requests (one request at a time)
- 10. Typical SolutionsWork Management ConcernsOverprovision resourcesSize thread pools for maximum load factor, usually requiring many load-test runs to optimizeWLS 8.1 - configure multiple execute queuesDeploy applications into different server instancesBuild additional machinery to detect overload condition and react=> Sub-optimal ROI and increased Complexity
- 11. WebLogic Server Work ManagersIndicate your intent
- 12. WebLogic is on autopilot adjusting to meet your goalsWebLogic Server Work ManagersCore PrinciplesWork PrioritizationApplications define resource requirements via meta-data they can relate to, rather than low-level technical constructs (thread counts)User-specific SLAs can be definedThread Pool ManagementApplications should not have to configure and maintain thread poolsWLS manages this internally and automaticallyWithout necessarily requiring Administrator configuration and sizing inputOverload ProtectionStandardized mechanism to respond to overload conditions
- 13. WLS Work ManagementKey ComponentsNew Thread Pool ImplementationSingle internally managed thread pool and priority-based request queue service all application requestsRequest “Priority” dynamic and internally computed to meet application-defined goalsThread Count Self-TuningSelf-tuning thread pool monitors overall throughput every two secondsPresent thread count, measured throughput, and past history determines if thread count needs to changeNew threads automatically added/removed as neededBenefits Administrators and Operators - no need to conduct tedious performance testing or guesswork just to pick a static thread pool size that does not adapt to changing workloads
- 14. WLS Work ManagementKey ComponentsWork ManagersRuntime abstraction used by applications to define resource requirementsWork Manager ComponentsRequest ClassFair-Share – desired share of server resources for appResponse Time –desired app response timeContext Based – user-specific SLAsMinimum Thread ConstraintMaximum Thread ConstraintCapacitySpecified in application descriptor (weblogic.xml, weblogic-ejb.xml, weblogic-application.xml)Can be accessed programmatically via CommonJ API – JSR-237
- 15. Work Manager ExamplesFair ShareDesired share of server resourcesThread usage become higher as fair share number increasesFair shares are relative to other fair shares defined in the system<work-manager><name>highfairshare_workmanager</name><fair-share-request-class> <name>high_fairshare</name> <fair-share>80</fair-share></fair-share-request-class> </work-manager>
- 16. Work Manager ExamplesResponse Time GoalDesired response-time goal in millisecondsResponse-time goals relative to other response goals and fair shares<work-manager><name>highfairshare_workmanager</name><response-time-request-class> <name>fast_response_time</name> <goal-ms>2000</goal-ms></response-time-request-class> </work-manager>
- 17. Work Manager ExamplesContext-BasedCurrently look at security name and group of user submitting the request<work-manager> <name>context_workmanager</name><context-request-class> <name>test_context</name> <context-case><user-name>platinum_user</user-name> <request-class-name>high_fairshare</request-class-name> </context-case> <context-case><user-name>evaluation_user</user-name> <request-class-name>low_fairshare</request-class-name> </context-case> </context-request-class></work-manager>
- 18. How to use Work ManagersCoarse Grained
- 19. Target at the entire server (“default” Work Manager)
- 20. Target entire applications or modules - e.g. weblogic.xml<wl-dispatch-policy>myAppWorkManager</wl-dispatch-policy>Fine Grained
- 21. Target individual JSPs, Servlets, EJBs, MDBs <servlet> ...<init-param> <param-name>wl-dispatch-policy</param-name> <param-value>myCustomWorkManager</param-value> ...Programmatically via JNDI lookupInitialContextic = new InitialContext();commonj.work.WorkManager wm = (commonj.work.WorkManager)ic.lookup("java:comp/env/wm/myWM");
- 22. Stuck ThreadsA Label given to threads not returned to thread pool after a configured period of time (defaults to 600 secs)
- 23. Does not mean threads are literally stuck or locked (although possible) – they might be performing a long-running task
- 24. This is an informational label – Administrators cannot stop the threads directly due to Java Threading design
- 25. The Server or the Work Managers can be configured to shutdown or go to an Admin State once a certain number of threads become stuck<Insert Picture Here>WLS Work Managers Demonstration
- 26. Demonstration ApplicationWeb Applicationnormal.jspveryslow.jsp – 30 second sleepWork Manager targeted at veryslow.jspMax Thread Constraint = 5Capacity Constraint = 7Test Scenario10 concurrent clients access veryslow.jsp (using JMeter)Work Manager controls how clients get serviced7 clients – OK3 clients – HTTP 503 (“Server Busy”)http://blogs.oracle.com/jamesbayer/entry/work_manager_leash_for_slow_js
- 27. WLS behavior under test scenario
- 28. Test Results
- 29. Image AttributionPhoto of Circuit Breaker courtesy of Alan Hochberg
Editor's Notes
- #14: So in this example, if another fair share was assigned a value of 20, the ration would be 80:20 or 4:1. This means that if each application has the same amount of requests coming in and there are not enough threads to server all requests immediately, then high-fair share work manager will get 4x as many requests serviced as the low-priority work manager.